JP2015219923A - Spam comment detection method and spam comment detection device - Google Patents

Abstract

PROBLEM TO BE SOLVED: To provide a method and device for detecting a spam comment on an image capture device.SOLUTION: A detection method includes the steps of: extracting a feature word and an evaluation word in a comment and determining the first evaluation polarity of the evaluation word; determining the second evaluation polarity of an image feature that corresponds to the feature word and evaluation word of a related image photographed by an image capture device; comparing the first and second evaluation polarities and, when the first and second evaluation polarities do not match, determining that the comment is a spam comment. The method has a relative high accuracy of detection due to that the objectivity of a comment is evaluated on the basis of a concrete fact such as comment information and a related image, and is able to detect a spam comment quickly due to that the method does not take much time and trouble.

Description

  The present invention generally relates to network information processing, and more particularly, to a spam comment detection method and apparatus for an image capture apparatus.

  In recent years, with the development of the Internet, people tend to express their views on the network. For example, when purchasing a product, the user can post comments on the product on various business sites, electronic bulletin boards, and blogs. Such comment information is extremely important for the merchandise provider, and the merchandise provider can contribute to improving the next generation of merchandise by grasping the defects and defects of the merchandise. On the other hand, such comment information is equally important for other users, and when other users quickly grasp the performance, advantages and disadvantages of the product and select the product, they will make a better purchase decision and selection Contribute.

  However, because of the openness of the network, people can write comments arbitrarily on the website, so the commentability of some comments is reduced, resulting in spam comments. For example, if the user thinks that the product does not satisfy the expectation without knowing the function of the product, the impression of the product may be deteriorated and the user may underestimate the product. As another example, a company employee may overestimate its product and underestimate a rival company's product. This less-objective spam comment has the negative impact on the product provider's inability to understand and improve the product ’s actual flaws, while other users can quickly understand the product ’s performance, benefits and drawbacks, And there is a risk of misunderstanding other potential users. Therefore, it is desirable to be able to determine a comment, which is a spam comment, and to eliminate interference with merchandise providers and other potential users.

  Conventionally, various studies have been made on how to detect spam comments. These considerations usually determine whether a comment is a spam comment by determining the objectivity of the comment based on text comment information and contributor information. For example, an ordinary spam comment detection technique first determines a poster's reliability based on a poster's history comment, and then determines whether the comment is a spam comment based on the reliability of each poster. However, according to this technique, it takes a lot of time to manually mark data such as a history comment of a poster, and the reliability of the poster determined by, for example, creating a probability model is high. Since it is not always reliable, the accuracy of detecting spam comments is relatively low.

  Embodiments of the present invention seek to provide a method and apparatus for detecting spam comments for an image capture device.

  According to one embodiment of the present invention, there is provided a spam comment detection method for an image capture device, the step of extracting a feature word and an evaluation word in a comment and determining a first evaluation polarity of the evaluation word; Determining a second evaluation polarity of an image feature corresponding to the feature word and the evaluation word of a related image captured by a capture device; comparing the first evaluation polarity with the second evaluation polarity; And a step of determining that the comment is a spam comment if the first evaluation polarity does not match the second evaluation polarity.

  According to another embodiment of the present invention, there is provided a spam comment detection apparatus for an image capture device, which extracts a feature word and an evaluation word in a comment and determines a first evaluation polarity of the evaluation word. A polarity determining means for determining a second evaluation polarity of an image feature corresponding to the feature word and the evaluation word of the related image photographed by the image capture device, and the first evaluation polarity and the second evaluation polarity. Comparing and providing a spam comment detecting device comprising: a comparing means for determining that the comment is a spam comment when the first evaluation polarity and the second evaluation polarity do not match.

  According to another embodiment of the present invention, there is provided a spam comment detection method for an image capture device, wherein for each opinion in a comment, a feature word and an evaluation word in the opinion are extracted, and the evaluation word A first evaluation polarity is determined, a second evaluation polarity of an image feature corresponding to the feature word and the evaluation word of the related image captured by the image capture device is determined, and the first evaluation polarity and the second evaluation polarity are determined. Comparing the polarities, and if the first evaluation polarity and the second evaluation polarity do not match, determining that the objectivity of the opinion is non-objective, and weighted averaging the objectivity of each opinion Calculating the objectivity of the comment and determining whether the comment is a spam comment, wherein the weighted average weight is determined based on the importance of each opinion; Detection method of Pam comment.

  According to the spam comment detection technique for the image capture device according to the embodiment of the present invention, whether or not the comment is a spam comment by determining the objectivity of the comment based on the match between the comment information and the related image. And the objectivity of the comment is evaluated based on specific facts, so that the detection accuracy is relatively high. In addition, the detection technique can quickly detect spam comments by eliminating a lot of time and effort.

2 illustrates an example of a comment and related images for an image capture device. 3 shows a flowchart of a spam comment detection method for an image capture device according to a first embodiment of the present invention. 6 illustrates comments and related images for another image capture device. Furthermore, the comment and related image about another image capture device are illustrated. The possible correspondence between the first evaluation polarity of the evaluation word in the comment and the second evaluation polarity of the image feature is shown. It is an illustration of camera parameter setting when the first evaluation polarity matches the second evaluation polarity and both are negative. 1 is a block diagram illustrating a function setting of a spam comment detection device for an image capture device according to an embodiment of the present invention. FIG. 1 shows an overall hardware block diagram of a spam comment detection system for an image capture device according to an embodiment of the present invention. FIG.

  In order that those skilled in the art may better understand the present invention, the present invention will be described in more detail below with reference to the drawings and specific embodiments.

  First, the basic purpose of the spam comment detection technique for the image capture device according to the embodiment of the present invention will be briefly described.

  When a user of an image capture device makes a comment on the image capture device, the related image captured by the image capture device is usually displayed simultaneously as a basis to prove the opinion in the comment. For example, as shown in FIG. 1, the camera user comments that the camera is not good, because the face appears dark when backlit. To prove this view, the user shows a picture taken with the camera in backlight. In view of such circumstances, in the spam comment detection technology according to the embodiment of the present invention, the comment is a spam comment based on the match between the comment and the corresponding image feature of the related image captured by the image capture device. It is determined whether or not. Specifically, if the comment does not match the quality of the corresponding image feature of the related image, the comment is non-objective and spam comment since it can be determined that the camera is over or underestimated. . On the other hand, if the comment matches the quality of the corresponding image feature and the image quality is poor, the cause of the poor image feature quality is to accurately set the shooting parameters of the image capture device when shooting the relevant image If the answer is YES, it can be determined that the comment of the image capture device is non-objective and is a spam comment.

  The image capture device may be various electronic devices such as a camera, a video camera, and a smartphone having a shooting function. Hereinafter, a camera will be described as an example.

<First embodiment>
FIG. 2 is a flowchart of a spam comment detection method for the image capture device according to the first embodiment of the present invention.

  As shown in FIG. 2, in step S201, a feature word and an evaluation word in the comment are extracted, and a first evaluation polarity of the evaluation word is determined.

  Evaluation words are words that express evaluation, and may be nouns, verbs, adjectives, exclamations and some idiomatic terms. Evaluation polarity reflects the tendency of evaluation and may include, for example, positive evaluation, neutral evaluation and negative evaluation. Evaluation words and evaluation polarities are both commonly used technical terms in this field, and detailed description thereof is omitted here.

Basically comment containing a single opinion one of the quintet _{O i = (e j, a} k, s l, h m, t n) (O i is contributor _{h m} were posted in time _{t n,} the target object / a opinion about _{a k} features / aspects of the entity _{e j, s l} can respectful with the evaluation value of the opinion _{O i.).} In the present embodiment, it is only necessary to extract a _k features / aspects and associated evaluation s _l of the target object / entity (camera). For example, it is necessary to extract the characteristic word “face” and the evaluation word “dark” for the above example comment “the camera is not good because the face appears dark when backlit”.

  Any known technique in this field can be used to extract the characteristic words and the evaluation words from the comments to determine the evaluation polarity of the evaluation words. For example, Liu, B.I. "Aspect-based Sentiment Analysis", Sentiment Analysis and Opinion Mining, 2012 (pages 49-78), Morgan & Claypool Publishers or Pang, B .; And Lee, L .; "Opinion Mining and Sentiment Analysis", Foundations and Trends in Information Retrieval, 2008 / January, page 1-135, can be used. In order to avoid unnecessary confusion of the inventive content of the present invention, the process is not described in detail here.

  In step S202, a second evaluation polarity of an image feature corresponding to the feature word and the evaluation word of the related image photographed by the image capture device is determined.

  As described above, when a user of an image capture device comments on the image capture device, the related image captured by the image capture device is usually displayed at the same time to prove the opinion in the comment. Therefore, the related image has an image feature corresponding to the feature word in the comment. In step S202, an image feature corresponding to the feature word in the comment is determined, and an evaluation polarity of the image feature of the related image is determined. Hereinafter, specific processing of this step will be described in detail.

First, it is necessary to determine which image feature is the image feature corresponding to the feature word and the evaluation word, which may be performed, for example, by searching a predetermined first mapping table. Table 1 shows an example of the first mapping table.

  Table 1 is preset based on the imaging principle and functions of an image capture device, for example, a camera. What is shown in Table 1 is only an example, and those skilled in the image processing field can easily extend this table. Also, according to current research, when users comment on product features / aspects, the feature words and evaluation words used tend to match, so a very long and comprehensive mapping table is prepared and all There is no need to cover comments, and the first mapping table can be reasonably set to cover most comments.

  After determining the image feature, the quality of the image feature of the related image can be detected, and the evaluation polarity of the image feature can be determined based thereon. For example, when the quality of the image feature of the related image is good, the evaluation polarity is positive, and when the quality of the image feature is poor, the evaluation polarity is negative and the quality of the image feature is normal. And the evaluation polarity is neutral.

  How to detect the quality of the image feature of the related image is known in the art, and those skilled in the art can detect the quality by various appropriate means according to the specific image feature.

  For example, it is assumed that the comment of the contributor is “photos taken indoors are always blurred” and shows the image shown in FIG. In this case, first, for example, by searching the mapping table, it is determined that the corresponding image feature is a luminance gradient value. Normally, the higher the luminance gradient value, the richer the edge information on the screen and the clearer the image. Therefore, when detecting the quality of the image feature, the average luminance gradient value of the image shown in FIG. 3 is calculated and compared with a determination threshold value determined by statistically calculating the luminance gradient values of a plurality of images. The quality can be determined. For example, if the calculated average brightness gradient value is less than 10, it is considered blurry, that is, the quality is poor, and if the average brightness gradient value exceeds 14, it is considered clear, that is, the quality is good and the average brightness If the slope value is 10-14, the sharpness is considered normal, i.e. the quality is normal.

  As another example, it is assumed that the comment of the contributor is “the color of the photograph taken by the camera is biased and always bluish” and indicates the image shown in FIG. In this case, first, it can be determined that the corresponding image feature is a color component. Therefore, when detecting the quality of image features, the RGB image shown in FIG. 4 is represented by CIEL * a * b * (L * is the luminance of the image, a * is the red / green component of the image, and b * is the yellow of the image. / Indicates the blue component.) Can be converted to space. In general, since the average value of the a * and b * components is far from the origin and the variance is relatively small in the color cast image, the average value and variance in the a * and b * components of the image shown in FIG. 4 are calculated. By doing so, it is possible to determine whether or not to cast a color, and it is possible to determine the quality of the image.

  It should be understood that both examples above are illustrative and do not limit the invention. Those skilled in the art can detect the quality by various appropriate means based on the specific image feature, and further determine the evaluation polarity based on the quality of the image feature.

  Returning to FIG. 2, in step S203, the first evaluation polarity and the second evaluation polarity are compared, and if the first evaluation polarity and the second evaluation polarity do not match, it is determined that the comment is a spam comment.

  For example, taking the image shown in FIG. 4 as an example, the evaluation polarity of a comment that “the color of the photograph taken by the camera is biased and always bluish” is negative, while as shown in FIG. By calculating the average value and variance of the color components in the image, it is detected that the quality of the color components in the image is normal, thereby determining that the evaluation polarity is neutral. In such a case, the first evaluation polarity and the second evaluation polarity do not match, that is, the comment about the camera does not match the quality of the corresponding image feature of the related image, thereby determining that the comment is non-objective. The comment is a spam comment because the contributor may deliberately or underestimate the camera just because he does not like the style of the image.

  On the other hand, comments about the camera may be better than the quality of the corresponding image feature of the related image. For example, the comment is “the photograph taken with the camera is colorful and beautiful”, that is, the first evaluation polarity is affirmative, and the color cast occurs in the image by detecting the image feature of the related image. The quality of the image feature is found to be poor, i.e. the second evaluation polarity is negative. At this time, since the first evaluation polarity and the second evaluation polarity do not match, and the poster may intentionally overestimate the camera only by personal preference, the comment is also a spam comment.

  The spam comment detection method of the image capture device according to the first embodiment of the present invention has been described above in detail. In this embodiment, a spam comment with low objectivity is detected by determining the objectivity of the comment based on the match between the comment information and the related image. The detection method has relatively high detection accuracy by determining objectivity of comments based on specific factors such as comments and related images rather than uncertain factors such as the reliability of the poster. Further, since the detection method does not require a large amount of labor, it is possible to quickly detect spam comments.

<Second embodiment>
In the spam comment detection method for the image capture device according to the first embodiment of the present invention, a comment having the same first evaluation polarity and the second evaluation polarity is silently recognized as an objective comment. Absent. Actually, when the first evaluation polarity and the second evaluation polarity match, the comment about the image capture device is not necessarily objective. For example, when the comment about the camera is negative and the quality of the image feature is relatively poor (that is, when the first evaluation polarity matches the second evaluation polarity and both are negative), the negative Image comments cannot be considered objective and the quality of the image features is relatively poor because it may be due to improper parameter settings when taking images with the camera, Do not assume that the camera is bad.

  Specifically, FIG. 5 shows a possible correspondence between the first evaluation polarity of the evaluation word in the comment and the second evaluation polarity of the image feature. As shown in FIG. 5, if both the first evaluation polarity and the second evaluation polarity are neutral or negative, it cannot be determined whether the corresponding comment is objective, The negative evaluation polarity may be due to improper parameter settings when taking an image with the camera, i.e., if the parameters are set correctly, the quality of the image features may be good, The second evaluation polarity may be positive.

  In view of the above circumstances, this embodiment provides an improved spam comment detection method for an image capture device. The spam comment detection method for the image capture device according to the present embodiment is substantially the same as the detection method described in the first embodiment, and the differences are the first evaluation polarity and the second evaluation polarity in step S203. After comparing the two, the comment having the same first evaluation polarity and the second evaluation polarity is further inspected to determine whether or not the comment is a spam comment. Specifically, in step S204, when the first evaluation polarity and the second evaluation polarity match and neither is positive, each shooting parameter is accurately set when shooting the related image with the image capture device. It is determined whether or not to set, and if at least one shooting parameter is not set correctly, it is determined that the comment is a spam comment. In this step S204, if the comment about the camera is affirmative and the quality of the image feature is poor (normal or bad), further inspection is performed. In this step, each relevant shooting parameter is inspected, and unless one of the shooting parameters is set correctly, the cause of the poor image quality is that the shooting parameters of the camera are not set correctly. By being able to determine, non-affirmative comments about the camera are non-objective and spam comments, and if each shooting parameter is set correctly, the quality of the image features is poor Is due to the performance of the camera, and the camera's non-positive comments are objective, not spam comments. Hereinafter, specific processing of this step will be described in detail.

First, the shooting parameters corresponding to the feature words and evaluation words in the comment must be determined, which may be done, for example, by searching a predetermined second mapping table. Table 2 shows an example of the second mapping table.

  Table 2 is preset based on the imaging principle and functions of an image capture device, for example, a camera. The camera shooting parameters shown in Table 2 are shooting parameters closely related to various negative / neutral evaluations arranged in the evaluation word column. That is, if these camera shooting parameters are not set accurately, the evaluation word column is shown. Invite various negative / neutral evaluations. What is shown in Table 2 is only an example, and those skilled in the image processing field can easily extend this table to cover more commentary situations.

  After determining the shooting parameters, the specific setting of each shooting parameter at the time of shooting is compared with the corresponding predetermined rational parameter setting to determine whether each shooting parameter is set correctly.

  The specific setting of each shooting parameter at the time of shooting can be acquired from the EXIF information of the image. As is well known, when an image is captured by an image capture device, for example, a digital camera, the digital camera collects a series of shooting parameter information in the shooting process and incorporates it into the EXIF of an image in, for example, a JPEG / TIFF format. EXIF includes a plurality of camera parameters set when an image is captured, such as an aperture, a shutter, a focal length, and a flash mode. Therefore, for example, if the shooting parameters to be inspected are determined by the second mapping table, specific settings of those shooting parameters at the time of shooting can be directly extracted from EXIF.

  By comparing the specific setting of the shooting parameter at the time of shooting with a predetermined rational parameter setting by any appropriate means, it can be determined whether or not the shooting parameter is set correctly. For example, the camera shooting parameters are divided into numerical shooting parameters and type selection shooting parameters according to the type, and for each numerical shooting parameter, the numerical value set at the time of shooting is set to a predetermined rational parameter setting. If it is the maximum / minimum achievable value of the image capture device set by, it is determined that the shooting parameter is set correctly, and each type selection type shooting parameter is set at the time of shooting. The type is a type set by a predetermined rational parameter setting, or the type set at the time of shooting is not the type set by the corresponding predetermined rational parameter setting, and the image capture device has the rational parameter setting If the type set by is not provided, it is determined that the shooting parameter is set correctly. Hereinafter, this process will be described in detail with reference to a specific example shown in FIG.

  FIG. 6 shows an example of camera parameter setting when the first evaluation polarity matches the second evaluation polarity and both are negative. For example, as shown in Example 1 of FIG. 6, the evaluation of the camera is negative that “the background blur is not sufficient”, the feature word is “background”, and the evaluation word is “not enough blur” "No", there are two related camera shooting parameters, namely, "aperture" and "focal length" of the digital type shooting parameters. As is well known in the art, in order to blur the background as much as possible, both the aperture and focal length should be set to their maximum values, i.e. the predetermined rational parameter setting is the "maximum value that the image capture device can achieve". is there. As shown in Example 1 in FIG. 6, the camera aperture range is F / 2.8-F / 5.6, and the focal length range is 28-560 mm (this value can be obtained from the camera manual, for example). At the time of shooting, the actual aperture setting is F / 3.2, and the actual focal length setting is 28 mm. At the time of shooting, it can be seen that the maximum aperture F / 2.8 that can be achieved is not set, that is, the aperture is not set accurately. As a result, the quality of the image feature shown in Example 1 of FIG. 6 “the contrast between the foreground and the background is not sufficiently high” is not caused by the camera itself, and the shooting parameters (aperture) of the camera are accurately set at the time of shooting. It can be seen that this is because the camera performance is not fully exhibited. Therefore, even if the first evaluation polarity of the evaluation word “blur is not enough” and the second evaluation polarity of “contrast” which is the image feature in the example 1 match (both are negative). Comments about the camera are non-objective and are spam comments.

  For example, as shown in Example 2 in FIG. 6, the evaluation of the camera is negative such that “the face appears dark when backlit”, the characteristic word is “face”, and the evaluation word is “ When it is “dark”, there are two related camera shooting parameters, that is, the type-type shooting parameters “photometry” and “flash”. As is well known in the art, spot metering should be employed and preferably the flash is turned on, i.e. predetermined rational parameter settings, to avoid the face appearing dark when there is insufficient light. Are “spot metering” and “(flash) on”. As shown in Example 2 in FIG. 6, the camera has four photometry modes of evaluation photometry, partial photometry, spot photometry, and center-weighted average photometry, and the flash is turned on and off in two modes (for example, camera The actual metering mode is set to evaluation metering and the actual flash is set to off at the time of shooting. It can be seen that shooting parameters such as flash and metering are not set correctly during shooting. Accordingly, the quality of the image feature “Face brightness is abnormal” shown in Example 2 of FIG. 6 is not caused by the camera itself, and the shooting parameters (photometry and flash) of the camera are accurately set at the time of shooting. It can be seen that this is because the camera performance is not fully exhibited. Therefore, in the first example, even if the first evaluation polarity of “dark” as the evaluation word matches the second evaluation polarity of “luminance” as the image feature (both are negative), the camera is informed. Comments are non-objective and are spam comments. However, in the case of Example 2, the camera has only three types of metering modes of evaluation metering, partial metering, and center-weighted average metering, no spot metering mode, and no flash is attached to the camera. Therefore, it is considered that the shooting parameters are set correctly, that is, the quality of the image feature "Face brightness is abnormal" is not caused by not setting the parameters correctly, but the function of the camera itself This is due to failure. Therefore, a negative rating for a camera is objective, i.e. not a spam comment.

  The spam comment detection method for the image capture device according to the second embodiment of the present invention has been described above, and mainly the differences between the second embodiment and the first embodiment have been described. In this example, if the comment about the camera is non-positive and the quality of the image feature is poor (negative or neutral), the non-positive comment is not directly recognized as objective. Further, it is detected whether or not the quality of the image feature is poor due to inappropriate parameter setting when an image is captured by the camera, thereby further improving the accuracy of detecting spam comments.

<Modification>
In each of the above-described two embodiments, it has been described as an example that only one opinion is included in one comment. However, the present invention can be applied not only to this case but also to a case where a comment includes a plurality of views such as “colorful and beautiful” and “background blur is not sufficient”. Specifically, when a plurality of views are included in one comment, the objectivity of each view is detected using the detection method according to the embodiment of the present invention for each view, and then each Whether or not the comment is a spam comment by assigning a weight to it based on the importance of the opinion and calculating the objectivity of the comments containing those opinions by weighted averaging the objectivity of each opinion To decide.

  In the above description, the evaluation tendency of the evaluation word is shown by the evaluation polarity, and the evaluation polarity is divided into positive evaluation, neutral evaluation, and negative evaluation. In fact, those skilled in the art further divide the evaluation polarity according to the difference in polarity intensity according to the actual demand. For example, the positive evaluation is further divided into a strong positive evaluation, a normal positive evaluation, and a weak positive evaluation, and the negative evaluation is further divided into a strong negative evaluation, a normal negative evaluation, and a weak negative evaluation. It is done.

<Overall Configuration of Spam Comment Detection Device for Image Capture Device>
FIG. 7 is a functional setting block diagram of the spam comment detection apparatus 700 for the image capture apparatus according to the embodiment of the present invention.

  As shown in FIG. 7, the spam comment detection device 700 extracts a feature word and an evaluation word in a comment and determines a first evaluation polarity of the evaluation word, and a related image taken by the image capture device. The polarity determination unit 720 that determines the second evaluation polarity of the image feature corresponding to the feature word and the evaluation word, the first evaluation polarity and the second evaluation polarity are compared, and the first evaluation polarity and the second evaluation polarity And the comparison unit 730 that determines that the comment is a spam comment.

  For the specific functions and operations of the extraction unit 710, the polarity determination unit 720, and the comparison unit 730, the related descriptions in FIGS. 2 to 6 can be referred to, and redundant descriptions are omitted here.

  When the first evaluation polarity and the second evaluation polarity match and both are affirmative, the detection apparatus 700 determines whether each imaging parameter is accurately set when the image capture apparatus captures a related image. A parameter detection unit 740 for determining If the parameter detection unit 740 determines that at least one shooting parameter is not set correctly, the comparison unit 730 determines that the comment is a spam comment. The specific functions and operations of the parameter detection unit 740 can refer to the related description of FIGS. 5 and 6 described above, and the redundant description is omitted here.

<System hardware configuration>
FIG. 8 shows an overall hardware block diagram of a spam comment detection system 800 for an image capture device according to an embodiment of the present invention. As shown in FIG. 8, the spam comment detection system 800 is used to input related images or information from the outside, for example, comments about the image capture device, related images, and the like. Or the like, and the spam comment detection method according to the embodiment of the present invention described above, or the spam comment detection apparatus according to the embodiment of the present invention. As a result of performing the above detection process outside the processing device 820 which may be another chip having capability, for example, it is used to output each comment with an objectivity (comment spam) mark, for example, An output device 830, which may be a display, a printer, etc., and the object tracking process. Data, for example, the first mapping table, the second mapping table, the first evaluation polarity, the second evaluation polarity, predetermined rational parameter setting, setting of each shooting parameter at the time of shooting, function description of the image capture device, etc. A storage device 840 that is used for non-volatile storage and may be a variety of volatile or non-volatile memory such as, for example, random access memory (RAM), read only memory (ROM), hardware, or semiconductor memory. And may be provided.

  The basic principle of the present invention has been described above with reference to specific embodiments. However, for those skilled in the art, all or any of the steps or components of the method and apparatus of the present invention may be performed on any computing device (including processing devices, storage media, etc.) or computing device network by hardware, firmware, software. Or a combination thereof, which can be realized by a person skilled in the art after reading the description of the present invention and using their basic programming skills.

  Therefore, the object of the present invention may be further realized by executing one program or a set of programs on any computing device. The computing device may be a known general purpose device. Therefore, the object of the present invention may be realized only by providing a program product including program code for realizing the method or apparatus. That is, such a program product also constitutes the present invention, and a storage medium storing such a program product also constitutes the present invention. Of course, the storage medium may be any known storage medium or any storage medium to be developed in the future.

  However, of course, in the apparatus and method of the present invention, each member or each step can be disassembled and / or reconfigured. Their disassembly and / or reconfiguration should be regarded as equivalent means of the present invention. In addition, the step of executing the above processing may be executed in order of time as naturally described, but is not limited thereto. Some steps may be performed in parallel or independently of each other.

  The specific embodiments described above do not limit the protection scope of the present invention. Those skilled in the art should understand that various modifications, combinations, subcombinations and alternatives can be made depending on design requirements and other factors. Any modifications, equivalent substitutions and improvements made without departing from the spirit and principle of the present invention shall fall within the protection scope of the present invention.

Claims (9)

A method of detecting spam comments for an image capture device,
Extracting a feature word and an evaluation word in a comment and determining a first evaluation polarity of the evaluation word;
Determining a second evaluation polarity of an image feature corresponding to the feature word and the evaluation word of the related image captured by the image capture device;
Comparing the first evaluation polarity with the second evaluation polarity, and determining that the comment is a spam comment if the first evaluation polarity and the second evaluation polarity do not match. Detection method. If the first evaluation polarity and the second evaluation polarity coincide with each other and are not positive evaluations, whether or not each imaging parameter is accurately set when the image capture device captures the related image A step of determining whether or not
The method according to claim 1, further comprising: determining that the comment is a spam comment if at least one of the shooting parameters is not set correctly. Determining a second evaluation polarity of an image feature corresponding to the feature word and the evaluation word of the related image taken by the image capture device;
Determining an image feature corresponding to the feature word and the evaluation word based on a predetermined first mapping table;
The method according to claim 1, further comprising: detecting a quality of the image feature of the related image and determining a second evaluation polarity of the image feature based on the quality. Determining whether each imaging parameter is set correctly when the image capture device captures the related image,
Determining each imaging parameter corresponding to the feature word and the evaluation word based on a predetermined second mapping table;
The spam comment detection according to claim 2, further comprising the step of comparing each shooting parameter set at the time of shooting with each predetermined rational parameter and determining whether each shooting parameter is set correctly. Method. The step of comparing each shooting parameter set at the time of shooting with each predetermined rational parameter and determining whether or not each shooting parameter is set correctly includes:
When the numerical value set when shooting the related image is the maximum value / minimum value achievable by the image capture device having a predetermined rational parameter, the shooting parameter is accurately set. The method of detecting a spam comment according to claim 4, further comprising the step of determining that it is. The step of comparing each shooting parameter set at the time of shooting with each predetermined rational parameter and determining whether or not each shooting parameter is set correctly includes:
For the type selection type shooting parameter, the type set when shooting the related image is a predetermined type of a predetermined rational parameter, or the type set when shooting the related image is a predetermined rational 6. The method according to claim 4 or 5, further comprising the step of determining that the shooting parameter is set correctly if the parameter is not a predetermined type of parameter and the image capture device does not include the predetermined type of rational parameter. Spam comment detection method described in. A spam comment detection device for an image capture device,
Extraction means for extracting a characteristic word and an evaluation word in a comment and determining a first evaluation polarity of the evaluation word;
Polarity determining means for determining a second evaluation polarity of an image feature corresponding to the feature word and the evaluation word of the related image captured by the image capture device;
Comparing means for comparing the first evaluation polarity with the second evaluation polarity, and determining that the comment is a spam comment if the first evaluation polarity and the second evaluation polarity do not match. Comment detection device. If the first evaluation polarity and the second evaluation polarity match and neither is a positive evaluation, whether or not each imaging parameter is correctly set when the image capture device captures the related image Further comprising parameter detection means for determining
The spam comment detection device according to claim 7, wherein the comparison unit determines that the comment is a spam comment when the parameter detection unit determines that at least one of the shooting parameters is not correctly set. . A method of detecting spam comments for an image capture device,
For each view in the comment,
Extracting characteristic words and evaluation words in the view, determining a first evaluation polarity of the evaluation words,
Determining a second evaluation polarity of an image feature corresponding to the feature word and the evaluation word of the related image captured by the image capture device;
Comparing the first evaluation polarity with the second evaluation polarity, and determining that the objectivity of the view is non-objective if the first evaluation polarity and the second evaluation polarity do not match; ,
Weighting the objectivity of each view, calculating the objectivity of the comment and determining whether the comment is a spam comment,
A method for detecting spam comments, wherein the weighted average weight is determined based on the importance of each opinion.

Images